OD5-04

INVESTIGATIONS OF IN-OIL METHODS FOR PD DETECTION AND

VIBRATION MEASUREMENT
1* 1 1
M. Beltle , M. Siegel , S. Tenbohlen
1
University of Stuttgart, Pfaffenwaldring 47, 70569 Stuttgart, Germany
*Email: michael.beltle@ieh.uni-stuttgart.de

Abstract: Usually, acoustic partial discharge (PD) and vibration measurements take
place on the outside tank wall of power transformers. This contribution determines new
approaches for acoustic PD measurements within the oil volume of the transformer
combined with electromagnetic in-oil measurements. Scope of the work is an acoustic
and ultra-high frequency (UHF) detection of PD and additional recording of transformer
vibrations during service. For in-oil measurements a combined sensor is presented, which
allows simultaneous detection of all three signals. After a short introduction of
measurement principles, the requirements of a combined in-oil sensor are defined
including mechanical demands and signal processing. A diplexer integrated into the
combined sensor is introduced to separate the considered frequency spectra. Tests in a
laboratory setup with a first prototype estimate if these requirements are met. Therefore,
the setup uses reproducible, synthesized PD signals generated by an acoustic signal
generator and an UHF pulser. Second measurements use a needle-sphere electrode
emitting realistic UHF and acoustic signals. Vibrations are simulated with a small winding,
which is mechanically excited. The combined sensor is compared to the UHF antennas
and acoustic accelerometers. Finally, the influence of the electric field strength being
unavoidable at inside-tank measurements is evaluated.

1 INTRODUCTION transformer tank with it’s electromagnetical

Reliable operation of power transformers is
essential for supply security. Damages to the
insulation of a power transformer, like local
defects, must be recognized at an early stage to
provide time for retaliatory action. Different
diagnostic methods have been established to meet
the deriving demands. Considered methods are
acoustic and ultra-high frequency (UHF) detection
of partial discharges and measurement of
transformer vibrations during service. Aim of this
contribution is to unite all needed sensor
technologies in one device.
1.1 Partial discharge measurement
The partial discharge (PD) measurement
represents standard procedure for the estimation
of local defects. PD are usually measured electrical
according to IEC 60270, which represents the
established method [1]. Also newer techniques, not
being normed yet, arise. Their essential advantage
is localisation of PD by run-time measurements.
1.1.1 UHF PD measurement
An Antenna inside the transformer measures the
electromagnetic emission of PD sources. PD
represent fast electrical processes radiating
electromagnetic waves with frequencies up to the
ultrahigh range (UHF: 300 – 3000 MHz) in the
surrounding oil [2], [3]. Due to the moderately
attenuated propagation of UHF waves inside the
transformer tank, the electromagnetic wave
detection is very sensitive. The grounded
shielding effect represents a Faraday cage.
Therefore the UHF measurement is less sensitive
to external noise like corona or other radiated
disturbances.
1.1.2 Acoustic PD measurement
Well-known acoustic effects of PD are audible at
corona on overhead lines. However, also internal
PD can emit acoustic signals in the ultrasonic
range [4]. Usually, the typical acoustic frequency
spectrum of internal PD reaches from 80 kHz to
180 kHz. Usually, accelerometers with a resonant
frequency of 150 kHz at the outside tank wall are
used for measurement. The characteristic noise
spectrum for power transformers has a frequency
range up to 80 kHz [5]. This is caused by
magnetostriction, external loose parts and also
rain, hail or wind-blown sand on the transformer.
The acoustic measurement is mostly used to
locate PD by run-time measurements. This is
complicated due to the structure borne path
problem [5]. Sonic speed in steel is faster than in
transformer oil. The superposition of all possible
signal paths measured from the external sensor
can lead to faulty run-time detection and therefore
erroneous PD localization.
1.2 Vibration measurement
Vibrations are caused by voltage-dependent and
load-dependent effects, which lead to oscillations
in mechanical structures of power transformers [6].
The voltage-dependent vibration is originated by
magnetostriction leading to oscillations of the core

798

(e.g. lamination sheets). The Weiss Domains in
metal align themselves along the time-varying
magnetic main flux induced by the applied voltage.
At load condition, current-related effects
superimpose magnetostriction. Forces of the
alternating magnetic field affect current-carrying
windings leading to an oscillation of the active part.
The basic oscillation for both phenomena is twice
the electrical frequency [7]. The entire frequency
range is usually small, reaching up to approx.
1 kHz [6].
2 COMBINED IN-OIL MEASUREMENT
Both, acoustic measurements for PD and
vibrations cope with the problem of the stiff, self-
supporting tank, which is influencing both, the
signal strength (damping) and the signal to noise
ratio (SNR). Therefore, methods for measurents
directly inside the transformer oil are determined.
Piezo-electric sensors represent a suitable
technology to detect changes of pressure caused
by acoustic waves. Their frequency range is suited
for vibration and PD acoustic measurements.
Unfortunately, the access to the transformers
inside is limited by the amount of available oil
valves. Most transformers are equipped with only
2-3 valves, which can be used for in-oil
measurements. Therefore, a sensor is required,
that is able to measure pressure changes by the
piezo-electric effect and electromagnetic waves
simultaneously. Thus, the sensor should work from
100 Hz up to the GHz range. Separate signal
outputs should be available for vibrations, acoustic
PD and UHF PD. The mechanic design of the
sensor has to be attachable to existing oil valves. A
prototype designed to meet the requirements is
shown in Figure 1. Compared to the combined
acoustic sensor presented in [8], this version is
also suited for UHF signals. A diplexer with three
output N-connectors is integrated in the handle of
presented sensor.
Figure 1: Combined sensor for UHF PD, acoustic
PD and vibration measurement. Right: sensor
head, middle: oil valve adaptor, left: three N-
connectors for separated signal outputs
2.1 Mechanical requirements
For the application to power transformers the
sensor is designed and tested to meet the
following demands
799
OD5-04
 mountable to DN 80 and DN 50 oil valves
 variable insertion depth, the sensor head
reaches into the transformer tank
 oil-proof, temperature resistant up to 80° C
continuous, 120° C short-term and
pressure resistant up to 5 bar
2.2 Principle of measurement
The block diagram in Figure 2 illustrates the mode
of operation. All signals, piezo-electric or UHF,
couple into the physical sensor unit. The UHF
signal is filtered by a passive high pass, providing
50 Ohms impedance up to 1 GHz. Because the
piezo-electric effect provides only very low output
voltage, the signal has to be amplified by a circuit
with high input impedance (in the GOhm range).
The amplifier also operates as low pass (LP) to
supress frequencies within the UHF range. The
amplified signals are separated by an active low
pass filter for vibrations. PD acoustics are filtered
by an active band pass (BP) ideally operating from
55 kHz to 200 kHz. Active filters are used due to
their amplification, smaller dimensions and better
pass band tuning.
Figure 2: Block diagram of diplexer to separate
signals from UHF PD, acoustic PD and sinusoidal
vibration signals to separate outputs. On the left:
sensor receiving all signals simultaneously
The transfer functions of the three separated signal
paths within the diplexer are shown in Figure 3.
Acoustic
Vibration PD
s
UHF
Figure 3: Transfer function of diplexer,
red: low pass filter for vibration measurement
blue: band-pass filter for acoustic PD (both active),
green: high pass filter for UHF PD (passive)

3 TEST MEASUREMENTS IN LABORATORY
SETUP
The prototype is tested in a small test tank filled
with mineral oil. Two types of tests are performed.
The first uses stable, reproducible synthetic signals
for vibrations, acoustic and UHF PD. The second
uses a needle-sphere electrode for realistic
acoustic and UHF PD signals. A sketch of the test
tank is shown in Figure 4.
Figure 4: Laboratory test setup for combined PD
and vibrations measurements
Vibrations are generated by a winding part, which
is mechanically excited by a low frequency
transducer. PD acoustic signals are inserted by a
tuneable signal generator. The measurement
setup is shown in Figure 5. A piezo-electric
transducer converts amplified, synthesized voltage
signals to in-oil sonic sound. As artificial UHF
source a LDC-7/UHF pulser from Doble Lemke is
used. Figure 4 shows the position of the combined
sensor and UHF antenna [9], used as reference.
Figure 5 illustrates the setup for PD acoustic and
vibration measurement. Accelerometers are used
for the outside tank wall of vibrations and PD
acoustics. In both cases, whose output is amplified
by 40 dB. All sonic signals are recorded in time
domain by a digital storage oscilloscope (DSO)
with 400 MHz analogue bandwidth. UHF signals
are recorded in time domain by a DSO with 3 GHz
bandwidth.
combined
sensor
DDS
synthesised
PD signal acceleration
sensor
source: (tankwall)
piezoelectric
transducer
Figure 5: Laboratory test setup for reproducible
acoustic PD measurements
OD5-04
The synthetic acoustic PD signal is shown in
Figure 6. The basic oscillating frequency is
adjustable as well as the rise time and the fall time.
To avoid interaction between two sequenced
signals, the signal repetition rate is set to 200 ms.
The position of the piezoelectric transducer is
static. Boundary effects like reflections on the tank
wall do not change during measurement. Using
this setup, the acoustic PD output of the combined
sensor is compared to the outside tank wall
acceleration sensor (resonance frequency
fres = 150 kHz).
Figure 6: PD signal generated by direct digital
synthesis (DDS) with variable parameters:
frequency, rise time and fall time (here: 180 kHz,
4 periods rise and 10 periods fall time)
3.1 Acoustic PD measurements
Acoustic PD measurements are mostly used for
localisation purposes by determining run-time
differences. Therefore, focus of the presented
analysis lies on the detectability of the starting time
of acoustic signals. All signals show single shot
measurements, no averaging for noise reduction is
used. The first measurements use the synthetic
signal setup with a basic oscillating frequency of
150 kHz. Simultaneously, the winding fragment is
excited with 100 Hz to simulate transformer
vibrations. Figure 7 shows the recorded acoustic
PD signals of both sensors.
The signals are triggered on the reference output
of the signal generator. The delays between signal
start and trigger are caused by the sonic speed in
oil of approx. 1400 m/s [10]. Both sensor output
DSO
signals are amplified by 40 dB and have a clearly
detectable start time (marked by the coloured
arrows at 250 µs for in-oil and 400 µs for tank-wall
measurement). The combined sensor has a
steeper derivation, which is advantageous for start
Ampilfier
40dB
time detection. The noise levels are approximately
equal. Both channels show multiple local maxima,
indicating reflections on the tank wall. Eventually,
both signals are well suited for start time detection.
800
 OD5-04

correct acoustic start time can be detected on both

sensors. Again, the SNR of the in oil measurement
is better than outside tank wall measurement,
whereas the absolute signal amplitude of the
accelerometer is higher in this case.

For real transformer measurements the in-oil

measurement indicates significant advantages due
to the fact, that the signal strength often lies below
the noise level of outside tank wall mounted
sensors, because of the too low SNR.

Figure 7: Signal of in-oil (blue) and outside tank

wall measurement (red). PD signal frequency
150 kHz

Figure 8 shows a measurement at an increased

basic oscillating frequency of 180 kHz. The piezo-
electric source has low sonic coupling at this
frequency resulting in smaller signal strength. The
coloured arrows mark the actual arrival time of PD
signals at the positions of the combined sensor
(blue) and accelerometer (red). It becomes
apparent, that the start time is still detectable in the
Figure 9: Acoustic measurement of needle-sphere
in-oil measurement, but not on the outside tank
PD source, triggered on UHF antenna. Low
wall. In contrast, higher amplitudes at 600 µs,
frequency coupling of electromagnetic components
which are caused by reflections, would indicate a
in combined sensor
delayed start time and hence result in an
erroneous PD source localisation. Considering the
SNR of both sensor locations, the in-oil 3.2 UHF PD measurements
measurement is also advantageous. The sensitivity of the combined sensor in the UHF
range is determined using a reproducible synthetic
signal and real UHF PD signals form the needle-
sphere electrode. An UHF antenna (also it is the
signal trigger) acts as reference. The setup is used
like shown in Figure 4. A single shot is shown in
Figure 10. The blue curve shows the output of the
passive UHF filter from the combined sensor and
the UHF antenna output (coloured in red).

Figure 8: Signal of in-oil (blue) and outside tank

wall measurement (red) at 40 dB amplification. PD
signal frequency 180 kHz

The results from the synthetic test setup are

confirmed using a needle-sphere electrode acting
as PD source. Figure 9 shows the sensor output
voltages recorded with a DSO. As trigger an Figure 10: UHF measurement of LDC-7 pulser.
additional UHF antenna is used. It becomes Red: UHF antenna, Blue: UHF diplexer output
apparent, that there is a small coupling of the UHF
signal into the acoustic PD path. This is tolerable
The distances between source and sensors are
because it does not affect the detectability of the
equal. The antenna shows nearly double levels at
acoustic start time. Due to the high level the
approx. double noise amplitude compared to the

801

combined sensor, thus their SNR is comparable.
With both sensors UHF triggering is possible. The
combined sensor can be disadvantageous at small
UHF amplitudes.Figure 11 shows the frequency
domain of the single shot. Red is the reference
from the UHF antenna. Direct comparison shows a
lower noise floor of the combined senor.
Figure 11: Frequency spectrum of time domain
measurement form Figure 10. Red: UHF antenna,
blue: UHF output of diplexer from combined sensor
Levels up to 400 MHz are comparable but then the
combined sensor’s sensitivity decreases with a
local minimum at 500 MHz. Antenna and sensor
differ approx. with 40 dB. For higher frequencies
the sensor is 10 dB below antenna level on
average (for noise and the second local maximum
at 900 MHz).Time domain measurement is also
performed with needle-sphere electrode to confirm
the results from the synthetic measurements, see
Figure 12.
Figure 12: UHF measurement of needle-sphere
electrode. Red: UHF antenna, Blue: UHF output of
diplexer from combined sensor
It can be stated, that the combined sensor is able
to detect UHF signals, but might be
disadvantageous if the UHF signal strength is very
low.
802
OD5-04
3.3 Vibration measurements
The evaluation of vibrations is also performed with
the setup described in Figure 4 and Figure 5. The
winding part is mechanically excited with different
frequencies being typical for vibrations: 100 Hz and
its harmonics. Measurements are performed in
time domain and transferred into frequency domain
using Fast Fourier Transformation (FFT). Figure 13
shows the frequency spectrum of the combined
sensor compared to an accelerometer mounted at
the outside tank wall. The winding is excited with
100 Hz and the signal is level measured (no
moving parts). Both sensors use 40 dB
amplification, thus the signals’ levels are
comparable. The combined sensor has higher
amplitudes with a very good SNR at basic
frequency showing 20 dBµV higher levels than the
accelerometer. Harmonics and sub-harmonics of
the basic frequency can be measured with both
sensors, which are caused by the mechanical
resonance frequency and stationary waves from
the tank geometry. The combined sensor shows a
significantly increased SNR.
Figure 13: Frequency spectrum of vibration
measurement at 100 Hz excitation.
Blue: in-oil measurement of combined sensor
Black: noise measurement in-oil
Red: Acceleration measurement outside tank wall
3.4 Influence of the electric field
Figure 14 shows a simulation of the electrical field
strength in the given setup geometry by a finite
element method solver. Maximum applied test
voltage is 70 kVRMS. The electric field strength at
the sensor position with used operation voltage for
PD generation in the measurement setup (30 kV)
and at a increased voltage of 70 kV are
determined. The value ERMS ~9 kV/m at an
operation voltage of 30 kV and ~22 kV/m for a test
voltage of 70 kV. For both cases the influence on
the output signals of the combined sensor is
determined. The UHF and acoustic PD outputs are
not influenced by 50 Hz due to high damping, see
Figure 3.

Figure 14: Simulation of the electrical field
strength in the test tank setup at applied voltage
VRMS = 70 kV
However, vibration measurement is influenced by
the electric field. Figure 15 shows the influence
compared to a noise measurement of the
combined sensor. At 50 Hz the effect of the electric
field results in an increased noise level. Because
50 Hz is usually not considered for vibration
analysis, the interference is acceptable. At 100 Hz
the maximum coupling of the electric field into the
vibration measurment path is max. 10dB above
noise and hence sufficiently low. Concluding, the
effect of the electric field on vibration measurement
is tolerable.
Figure 15: Frequency spectrum comparing
vibration measurement and influence of the electric
field. blue: noise measurement of combined sensor
green: coupling at ERMS ~9 kV/m,
orange: coupling at ERMS ~22 kV/m
4 CONCLUSION
The contribution presents a prototype of a
combined sensor being able to measure
transformer vibrations, PD acoustics and UHF PD
signals simultaneously. Signals are separated by
an included frequency diplexer. The combined
sensor is designed for application inside power
transformers. Laboratory tests evaluate proper
operation of the combined sensor. Its abilities and
limits compared to conventional sensors are
determined. The combined sensor is
advantageous to conventional techniques
concerning signal quality and SNR. Outside tank
803
OD5-04
wall sensors suffer signal damping by the
structure. The sensitivity of UHF of the combined
sensor is slightly smaller in comparison to
dedicated antennas. It is considered suitable for
simultaneous threefold measurements. Tests on
power transformers will have to evaluate its
usability in practice.
5 REFERENCES
i
[1] International Electrotechnical Commission,
IEC 60270 High-Voltage Ttest Techniques -
Partial Discharge Measurements, 3. Edition,
Geneva, Switzerland, 2000.
[2] S. Coenen, S. Tenbohlen, S. Markalous, T.
Strehl, „Sensitivity of UHF PD Measurements
in Power Transformers,“ in IEEE Transactions
on Dielectrics and Electrical Insulation, Vol.
15, , No. 6, 2008.
[3] M. D. Judd, „Experience with UHF Partial
Discharge Detection and Location in Power
Transformers,“ in Electrical Insulation
Conference, Maryland, USA, 2011.
[4] E. Howells, E. Norton, „Detection of Partial
Discharges in Transformers Using Acoustic
Emission Techniques,“ IEEE Transactions of
Power Apparatus and Systems, PAS-97,
1997.
[5] S. Markalous, Detection and Location of
Partial Discharges in Power Transformers
using acoustic and electromagnetic signals,
Stuttgart: Sierke Verlag, 2006.
[6] M. Beltle, S. Tenbohlen, „Usability of Vibration
Measurement for Power Transformer
Diagnosis and Monitoring,“ in International
Conference on Condition Monitoring and
Diagnosis (CMD), Bali, Indonesia, 2012.
[7] B. García, J. Burgo, and Á. Alonso,
„Transformer Tank Vibration Modeling as a
Method of Detecting Winding Deformations
Part I: Theoretical Foundation,“ IEEE
Transactions on Power Delivery, Vol. 21, No.
1, pp. 157-163, January 2006.
[8] M. Beltle, M. Siegel, S. Tenbohlen, „Analysis
of Methods for PD-Detection and Vibration
Measurement,“ in Diagnostik Elektrischer
Betriebsmittel, Fulda, Germany, 15-16
November, 2012.
[9] S. Coenen, S. Tenbohlen, T. Strehl, S.
Markalous, „Fundamental Characteristics of
UHF PD Probes and the Radiation Behavior of
PD Sources in Power Transformers,“ in
International Symposium on High Voltage
Engineering (ISH), Cape Town, South Africa,
2009.
[10] E. Howells, E.T. Norton, "Parameters Affecting
the Velocity of Sound in Transformer Oil,"
IEEE Transactions on Power Apparatus and
Systems, PAS-103, No.5, pp. 1111 -1115,
May 1984.